Graceful Departure

I first mentioned the idea of graceful departure in the node aging thread. What do people think of this idea?

Background

If a node leaves the network, there’s a need to secure the data they cared for as soon as possible, before other nodes leave and quorum may be lost. This rearrangement comes at the cost to adding new data to the network; new data must wait for the existing data to be secured.

Motivation

The aim is to reduce network load when nodes leave the network.

Possible implementation

The way this might happen is nodes may announce their imminent departure to the rest of the group, eg ‘I will leave this group in 75s’. This allows the other nodes in the group to prepare for this event using low-priority messages to try to redistribute that data before the node departs.

If the group doesn’t get there in time, or the node departs earlier than expected, or doesn’t announce departure at all, the rearrangement shifts into high priority mode.

I think this would be a very advantageous feature, not high priority on the development list, but certainly one to consider.

6 Likes

Reward for good behaviour is great. This may help but we do probably need to closely define the “roles” nodes play and what we need form them in more detail. Node ageing and data chains play a big role here so will alter the role somewhat (for the better). This is exactly though the reason for the RFC process so this would be a good one to consider.

5 Likes

Makes sense about the RFCs.

I think the ideal behavior is exactly enough downtime for reboots. The more we incentivize good behavior the closer we can get to exactly that. I think this problem can be partially addressed by selling hardware that will ensure that (except in cases where ISPs assign private address ranges) NAT and its associated hassles are minimized. Teredo ipv6 further boosts the availability of even natted nodes.

We had a realization today that no network of nodes that is sufficiently distributed like this exists and that the topology/topography/cartography/ontology of the actual Internet as it actually is verified on a blockchain did not exist. This also prevents “IoT” from turning into Internet of shit by allowing us to produce a verifiable history of sensor readings, that is controlled by the user. (their node can of course submit them as encrypted or world readable or paid readable). And because the devices connect to several distributed systems, the devices can submit on several systems as well. In the end you get a market solution to not losing every person’s privacy forever (thanks Jaron Lanier).

Hardware stands a reasonable shot at reducing the software complexity needed to get maidsafe (not to mention many other highly diffuse projects) up and running.

It might be high priority… It may be easy to implement and provide an outsized reward.